From ten blue links to AI answers: how search stopped sending clicks

For twenty years, the deal behind search was simple. You published a useful page. Google showed it as one of ten blue links. People clicked, and you got the visit.

That deal is ending, and not suddenly. Over 14 years, Google and then ChatGPT moved the answer, step by step, off your website and onto their own page. When Google shows an AI summary today, people click an ordinary result on 8% of visits, against 15% when there is no summary, according to Pew Research Center.

This is the story of how that happened, in seven chapters, and what a marketer or founder should do about it.

Things, not strings

In May 2012, Google launched the Knowledge Graph. The company described it as a shift from matching "strings" of text to understanding "things": people, places and organisations, and how they connect. At launch it held more than 500 million objects and 3.5 billion facts about them.

The visible result was a box on the right of the results page. Search for a famous person or a city and you got their key facts, with no click needed.

Few marketers worried at the time. The panel answered trivia, not the questions businesses compete on. But the principle was set. Google would now answer some questions itself, using facts it had gathered from across the web.

Lesson: when a platform starts storing facts about your category, how it describes you matters as much as where you rank.

The snippet that answered for you

In 2014, Google began showing featured snippets: a paragraph lifted from one web page and placed above the normal results, with a link to the source.

For publishers, snippets were a mixed gift. Winning one put you at the very top of the page. But if the snippet held the whole answer, the reader often had no reason to click. The "position zero" race began, and SEO teams started writing short, quotable definitions just below their headings.

This was the first time many marketers met the tension that defines search today. Being the source of the answer and getting the visit are not the same thing.

Lesson: pages written to be quoted get quoted. Put a clear, short answer near the top of every important page.

The zero-click years

By the early 2020s, a large share of searches ended on Google itself. Weather, sports scores, conversions, definitions, local business hours and flight times were all answered on the page.

SparkToro's 2024 study, based on clickstream data from Datos, put numbers on it. In the US, 58.5% of Google searches ended without a click to any result. For every 1,000 US searches, only 360 clicks went to the open web.

Two other changes made Google better at answering. In 2019 it added BERT, a language model that it said would help it understand about one in ten English searches in the US better. Long, conversational questions began to work. People started typing whole sentences.

Lesson: "traffic down, rankings stable" is often not a ranking problem. It can be the results page answering the question for you.

The chatbot that skipped the results page

On 30 November 2022, OpenAI released ChatGPT. It had no results page at all. You asked, it answered, in full sentences.

Growth was fast. By October 2025, Sam Altman said more than 800 million people used ChatGPT every week, up from 500 million at the end of March 2025.

ChatGPT changed what people expected from a question. Instead of ten options to compare, they got one confident answer, often naming specific brands. When ChatGPT added web search, it also began citing sources, but the citation sat below a complete answer.

For brands, this was a new kind of visibility. There was no position one, only whether you were named and how you were described.

Lesson: in an AI answer there is no page two. You are either in the answer or you are not.

Google answers back

Google's response came in stages. In 2023 it tested the Search Generative Experience with opted-in users. In May 2024, at its I/O conference, it launched AI Overviews to everyone in the US: an AI-written summary at the top of the results, with links to sources.

Adoption was fast because Google did not ask anyone to change habits. The summary simply appeared on searches people were already doing. By the first quarter of 2025, Google said AI Overviews reached 1.5 billion users a month, and by the second quarter, more than 2 billion.

In 2025, Google went further with AI Mode, a separate tab where the whole experience is a conversation, closer to ChatGPT than to classic search. By August 2026, Alphabet said AI Mode had passed 1 billion monthly users.

Lesson: the biggest change in search did not need people to switch tools. It came to the tool they already use.

The click cost

What did this do to clicks? The clearest independent evidence comes from Pew Research Center. It studied the browsing of 900 US adults in March 2025, covering 68,879 Google searches.

When an AI summary appeared, people clicked an ordinary search result on 8% of visits. Without a summary, they clicked on 15%. They clicked a link inside the summary itself on just 1% of visits. And they ended their browsing session after 26% of pages with a summary, against 16% without one.

That is close to half the clicks gone on the searches where a summary appears. The answer is still built from websites. The visit is not.

Lesson: measure whether you are named and cited in AI answers, not only how many visits search sends you.

What this means for you

The old deal of a page, a ranking and a click has not disappeared, but it now covers a shrinking share of questions. The new deal has three parts.

First, be in the answer. AI systems build their answers from pages they can find and trust. Clear, factual pages on your own site, with the answer near the top, are still the base. Then make sure the third-party sites AI systems read, such as review sites, comparison articles and industry press, describe you accurately.

Second, track the answer, not just the ranking. Look at the questions your buyers ask ChatGPT, Gemini and Google's AI Mode. Check whether you are named, in what position, next to which competitors, and which sources are cited. This is what tools like Aiso measure.

Third, make every visit count. If fewer people click through, the ones who do arrive further along, often after an AI has already compared options for them. Answer their remaining questions straight away and capture the lead while they are there. An AI chat agent such as Javo does that on your own site.

Lesson: visibility now happens in the answer, and conversion happens in the fewer, warmer visits that follow.